A New Method for Identification of Broadband Oscillations Based on ICEEMDAN and Fusion Classification Model

doi:10.11871/jfdc.issn.2096-742X.2025.02.016

Abstract

Abstract:

[Objective] To enhance the accuracy of wideband oscillation identification in power systems, this paper proposes a method that integrates ICEEMDAN-KPCA-KAN for wideband oscillation identification. [Methods] Firstly, for the original sequence with strong randomness and volatility, we use ICEEMDAN to decompose the original wideband oscillation signal into multiple intrinsic mode functions (IMFs) and a residual component, thereby reducing modal aliasing and highlighting the time-frequency characteristics of each component. Secondly, For each IMF, we extract time and frequency domain features and determine whether to use the IMF itself or its time-frequency features based on correlation. IMFs used directly are classified using a combined CNN-LSTM model. For those that need time-frequency features, we use KPCA to reduce the dimensions of high-dimensional data and create a feature set. This reduced data is fed into the Kolmogorov-Arnold Network (KAN) for classification. Finally, a LightGBM model combines the outputs from both classifiers to produce the final classification result. [Conclusions] Simulation examples demonstrate that this integrated method can quickly and accurately identify wideband oscillation signals with an identification rate of 99.73%.

Key words: wide-band oscillation identification, ICEEMDAN, KPCA, Kolmogorov-Arnold

LI Pengbo,ZHU Xiaofeng,SHEN Mingkang,SHA Haoyuan,HE Maohui,DENG Kai,XU Jin. A New Method for Identification of Broadband Oscillations Based on ICEEMDAN and Fusion Classification Model[J]. Frontiers of Data and Computing, 2025, 7(2): 161-174, https://cstr.cn/32002.14.jfdc.CN10-1649/TP.2025.02.016.

Figures/Tables 16

Table 1

Time domain characteristic parameters"

时域	公式	时域	公式
均值	$X - = 1 N ∑ i = 1 N x i$	最大值	$X m a x = m a x x i$
方差	$σ 2 = 1 N - 1 ∑ i = 1 N x i - X - 2$	波形因子	$C b = X R M S X$
均方根	$X R M S = 1 N ∑ i = 1 N x i 2$	峰值因子	$I p = X m a x X R M S$
方根幅值	$X R = 1 N ∑ i = 1 N x i 2$	脉冲因子	$C f = X m a x X$
绝对平均幅值	$X ¯ = 1 N ∑ i = 1 N x i$	裕度因子	$C e = X m a x X R$
偏斜度	$S i = 1 N ∑ i = 1 N x i - X - 3$	偏斜度因子	$P i = 1 N ∑ i = 1 N x i - X - 3 X R M S 3$
峭度	$β = 1 N ∑ i = 1 N x i 4$	峭度因子	$C q = 1 N ∑ i = 1 N x i - X - 4 X R M S 4$

Table 1

Table 2

Frequency domain characteristic parameters"

频域	公式	频域	公式
1	$F F 1 = 1 K ∑ k = 1 K s k$	6	$F F 6 = ∑ k = 1 K f k - F F 5 2 s k K$
2	$F F 2 = 1 K - 1 ∑ k = 1 K s k - F F 1 2$	7	$F F 7 = ∑ k = 1 K f k 2 s k ∑ k = 1 K s k$
3	$F F 3 = ∑ k = 1 K s k - F F 1 3 K F F 2 3$	8	$F F 8 = ∑ k = 1 K f k 4 s k ∑ k = 1 K f k 2 s k$
4	$F F 4 = ∑ k = 1 K s k - F F 1 4 K F F 2 2$	9	$F F 9 = ∑ k = 1 K f k 2 s k ∑ k = 1 K s k ∑ k = 1 K f k 4 s k$
5	$F F 5 = ∑ k = 1 K f k s k ∑ k = 1 K s k$	10	$F F 10 = F F 6 F F 5$

Table 2

Fig.1

Fig.2

Fig.3

Table 3

Fig.4

Fig.5

Table 4

Table 5

Fig.6

Fig.7

Fig.8

Fig.9

Table 6

Table 7

References 23

[1]	张东辉, 陈新. 基于阻抗视角的新能源发电系统宽频振荡抑制技术综述[J/OL]. 中国电机工程学报, 1-20[2024-07-19]. https://doi.org/10.13334/j.0258-8013.pcsee.231365.
[2]	BLAABJERG F, YANG Y, KIM K A, et al. Power electronics technology for large-scale renewable energy generation[J]. Proceedings of the IEEE, 2023, 111(4): 335-355.
[3]	马宁宁. 高比例新能源和电力电子设备电力系统的宽频振荡研究综述[J]. 中国电机工程学报, 2020, 40(15): 4720-4732.
[4]	王渝红, 王馨瑶, 廖建权, 等. 交直流电网宽频振荡产生、辨识及抑制研究综述[J]. 高电压技术, 2023, 49(8): 3148-3162.
[5]	LI N, HE F, MA W, et al. The identification of ecg signals using wavelet transform and woa-pnn[J]. Sensors, 2022, 22(12): 4343.
[6]	唐塬, 常俊杰, 彭少雄, 等. 基于EMD与神经网络的超声栓子信号分类研究[J]. 中国测试, 2023, 49(8): 15-20.
[7]	LU X, LU Z, WU Q, et al. Soft Fault Diagnosis of Analog Circuit Based on EEMD and Improved MF-DFA[J]. Electronics, 2022, 12(1): 114.
[8]	王建元, 张少锋. 基于线性判别分析和密度峰值聚类的异常用电模式检测[J]. 电力系统自动化, 2022, 46(5): 87-98.
[9]	马永光, 冯勇升. 基于IICEEMDAN-PCA-GRU的风电机组齿轮箱故障预警方法研究[J]. 太阳能学报, 2023, 44(4): 67-73.
[10]	欧阳静, 秦龙, 王坚锋, 等. 基于PCA-ShapeDTW-QWGRU的分布式光伏集群短期功率预测[J]. 太阳能学报, 2024, 45(5): 458-467.
[11]	王昱皓, 武建文, 马速良, 等. 基于核主成分分析-SoftMax的高压断路器机械故障诊断技术研究[J]. 电工技术学报, 2020, 35(S1): 267-276.
[12]	WU C, CHEN H, DU B, et al. Unsupervised Change Detection in Multitemporal VHR Images Based on Deep Kernel PCA Convolutional Mapping Network[J]. IEEE Transactions on Cybernetics, 2022, 52(11): 12084-12098.
[13]	李丹, 杨保华, 张远航, 等. 基于最优RBF核主成分的空间多维风电功率降维及重构[J]. 电网技术, 2020, 44(12): 4539-4546.
[14]	LIANG Y, LIN Y, LU Q. Forecasting gold price using a novel hybrid model with ICEEMDAN and LSTM-CNN-CBAM[J]. Expert Systems with Applications, 2022, 206: 117847.
[15]	王淑青, 盛世龙, 王云鹤, 等. 基于ICEEMDAN-VMD与结合注意力机制的简单循环单元的水轮机故障诊断[J/OL]. 中国农村水利水电, 1-16[2024-07-19]. http://kns.cnki.net/kcms/detail/42.1419.tv.20240514.1034.012.html.
[16]	LI T, QIAN Z, HE T. Short-Term Load Forecasting with Improved CEEMDAN and GWO-Based Multiple Kernel ELM[J]. Complexity, 2020, 2020: 1-20.
[17]	王双海, 米大斌, 芦浩, 等. 基于MIGA-VMD和t-SNE的轴承故障诊断方法[J]. 机床与液压, 2022, 50(2): 183-191.
[18]	纪德洋, 金锋, 冬雷, 等. 基于皮尔逊相关系数的光伏电站数据修复[J]. 中国电机工程学报, 2022, 42(4): 1514-1523.
[19]	付铭, 朱明, 梅杰, 等. 基于多元振动序列共性特征的电抗器故障诊断[J/OL]. 电测与仪表, 1-13[2024-07-19]. http://kns.cnki.net/kcms/detail/23.1202.TH.20240522.1259.002.html.
[20]	黄思航. 基于KPCA-EMD预处理方法的顶管机齿轮箱故障诊断方法研究[D]. 合肥: 合肥工业大学, 2023.
[21]	MCKELVY J A, NOVIKOVA I, MIKHAILOV E E, et al. Application of kernel principal component analysis for optical vector atomic magnetometry[J]. Machine Learning: Science and Technology, 2023, 4(4): 045048.
[22]	SU X, ZHANG H, ZHANG J, et al. Kernel-PCA-based single-phase earth fault detection model using multilayer perceptron in deep learning[J]. IET Generation, Transmission & Distribution, 2024, 18(4): 834-843.
[23]	宋匡玮, 吴浩, 陈伟哲. 基于CNN-LSTM网络的交直流电网故障线路识别方法[J]. 四川轻化工大学学报(自然科学版), 2024, 37(5) : 50-58.

振荡频率	类别
10~40 hz	0
70~100 hz	1
130~160 hz	2

IMF分量	与原始信号相关系数	P值
IMF1	0.44	0.02
IMF2	0.32	0.03
IMF3	0.45	<0.01
IMF4	0.74	<0.01
IMF5	0.34	<0.01
IMF6	0.08	0.02
IMF7	0.15	<0.01

信号	时频域特征与信号的相关系数	选取特征
原始信号	0.15	原始信号时频域参数
IMF4	0.12	IMF4时频域参数
IMF3	0.07	IMF3
IMF1	0.11	IMF1时频域参数
IMF5	0.04	IMF5
IMF2	0.01	IMF2

辨识模型	准确率/%	精确率/%	召回率/%	F1-score /%
EMD-KPCA-KAN	86.76	86.75	86.75	86.44
EEMD-KPCA-KAN	90.74	90.53	90.76	90.89
CEEMDAN-KPCA-KAN	92.89	92.84	92.76	92.80
ICEEMDAN-PCA-KAN	93.52	93.17	93.05	93.28
ICEEMDAN-CNN-LSTM	91.46	90.57	90.43	90.77
ICEEMDAN-TCN-BiLSTM	92.36	91.38	92.60	92.21
ICEEMDAN-KPCA-ELM	96.59	95.42	95.98	96.32
ICEEMDAN-KPCA-XGBoost	97.66	97.56	97.33	97.61
本文模型	99.73	99.49	99.51	99.46